An edge-preserving adaptive image denoising

作者：Wang, Xiang-Yang^*; Liu, Yang-Cheng; Zhang, Na; Wu, Chang-Jian; Yang, Hong-Ying

来源：Multimedia Tools and Applications, 2015, 74(24): 11703-11720.

DOI：10.1007/s11042-014-2258-x

• Image denoising
• Nonsubsampled shearlet transform
• Fuzzy support vector machines
• Adaptive Bayesian threshold

Based on nonsubsampled shearlet transform (NSST) and fuzzy support vector machines (FSVMs), we present a new denoising approach that can effectively suppress noise from an image while keeping its features intact. The noisy image is firstly decomposed into different subbands of frequency and orientation responses using NSST. The NSST detail coefficients are then divided into edge/texture-related coefficients and noise-related ones by FSVMs classifier. And finally the detail subbands of NSST coefficients are denoised by using the adaptive Bayesian threshold. Extensive experimental results demonstrate that our approach is competitive relative to many state-of-the-art denoising techniques. Especially, the proposed method can preserve edges very well while removing noise.

• 出版日期2015-12
• 单位辽宁师范大学; 苏州大学